STUDY ON SEGREGATION BEHAVIOR OF ALLOYING ELEMENTS IN TITANIUM ALLOYS DURING SOLIDIFICATION

STUDY ON SEGREGATION BEHAVIOR OF ALLOYING ELEMENTS IN TITANIUM ALLOYS DURING SOLIDIFICATION AKIRA KAWAKAMI B.E., Kyushu University (Japan), 1987 M.E., Kyushu University (Japan), 1989 A THESIS SUBMITTED IN PARTIAL FULFUFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES (Department of Metals and Materials Engineering) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA May 2002 Akira Kawakami, 2002

In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, 1 agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of The University of British Vancouver, Canada Columbia Date 0T/Z2. /OZ. DE-6 (2/88)

ABSTRACT A fundamental study was conducted on segregation behavior of alloying elements in titanium alloys to clarify the formation mechanism of "beta-flecks", melt-related defects enriched in beta stabilizing elements, which can cause a decrease in mechanical performance. Commercial titanium alloys, which are prone to the beta-fleck formation, such as 10-2-3 (Ti-10%V-2%Fe-3%Al), Ti-17(Ti-5%Al-2%Sn-2%Zr-4%Mo-4%Cr) and 6242 (Ti-6%Al-2%Sn-4%Zr-2%Mo) were used. Solidification parameters, such as dendrite arm spacing, distribution coefficients and densities of solid/liquid phase during solidification, were investigated in these alloys. Electron Probe Micro Analysis (EPMA) revealed that periodicity in distribution profiles of alloying elements corresponded to the primary or secondary dendrite arm spacing both in laboratory melted 10-2-3 ingots and in production 10-2-3 and Ti-17 ingots. This result indicates that EPMA is an effective method to clarify the dimensions of dendrite structures in titanium alloys(no good etching technique has been demonstrated that resolves the original dendritic structure). Distribution coefficients of alloying elements in 10-2-3, Ti-17 and 6242 were experimentally obtained using a zone-melting furnace. Distribution coefficients for iron in 10-2-3, zirconium and molybdenum in 6242 were deviated from the equilibrium distribution coefficients calculated from the binary phase diagrams. The fraction solidified (fs) at the initiation of beta-flecks was estimated to be 0.9 in 10-2-3 and Ti-17 using the Scheil equation, in which experimentally obtained distribution coefficients were used. The density of liquid and solid metal at around the melting point was estimated with the calculation software "METALS" and it was clarified that solid metal is heavier than liquid enriched with iron in 10-2-3 and that enriched with ii

chromium in Ti-17. The Rayleigh number was calculated to exceed 1 when the periodicity of chromium observed in a Ti-17 production ingot was assumed to be primary dendrite arm spacing. This fact suggests that the density-driven downward flow of liquid metal can occur. This in turn could cause channels perpendicular to the solidification direction and lead to the formation of beta-flecks, and supports the proposed mechanism. However, there are still some questions about the mechanism, such as the possibility of fluid flow at the final stage of solidification and the validity of considering the periodicity as primary dendrite arm spacing. iii

Table of Contents Abstract Table of Contents List of Tables List of Figures List of Symbols Acknowledgements ii iv vi vii ix xi 1. INTRODUCTION 1 2. LITERATURE REVIEW 3 2.1 SEGREGATION BEHAVIOR OF ALLOYING ELEMENTS IN TITANIUM ALLOYS(GENERAL) 3 Solidification microstructures in titanium alloys 3 Dendrite arm spacing 4 Solidification parameters 6 Distribution coefficients 6 Solidus and liquidus temperature 8 Solid state diffusion 8 2.2 BETA-FLECKS IN TITANIUM ALLOYS 10 Features of beta-flecks 10 Effects of beta-flecks on mechanical properties 13 Features of freckles 15 Formation mechanisms of freckles and beta-flecks 17 Freckle criterion and its application to beta-flecks 19 2.3 SUMMARY OF LITERATURE REVIEW 22 3. RESEARCH OBJECTIVES 34 4. EXPERIMENTAL METHODOLOGY 36 4.1 CHOICE OF ALLOYS 36 4.2 EXPERIMENTAL METHODS 37 Segregation behavior of iron in 10-2-3 small ingots melted and cast in an argon arc melting furnace 37 Segregation behavior of alloying elements in production ingots 38 Segregation behavior of alloying elements in laboratory melted small ingots using a zone melting furnace 39 iv

5. EXPERIMENTAL RESULTS 43 5.1 SEGREGATION BEHAVIOR OF ALLOYING ELEMENTS IN TITANIUM ALLOY INGOTS SOLIDIFIED IN DENDRITIC MANNER 43 Segregation behavior of iron in laboratory melted 10-2-3 ingots using an argon arc furnace 43 Segregation behavior of alloying elements in production ingots 45 5.2 SEGREGATION BEHAVIOR OF ALLOYING ELEMENTS IN ZONE MELTED TITANIUM ALLOY INGOTS 48 5.3 DENSITY OF BETA-FLECK AND LIQUID METAL CALCULATED USING "METALS" FOR 10-2-3 AND Ti-17 ALLOYS 54 6. DISCUSSION 74 6.1 DETERMINATION OF DENDRITE ARM SPACING BY EPMA 74 The relationship between dendrite arm spacing and solidification conditions in laboratory-melted ingots using argon an arc furnace 74 Dendrite arm spacing in production ingots 75 6.2 SEGREGATION COEFFICIENTS OF ALLOYING ELEMENTS IN COMMERCIAL TITANIUM ALLOYS AND ESTIMATION OF THE FRACTION SOLIDIFIED AT THE INITIATION OF BETA-FLECKS 77 Segregation coefficients of alloying elements in zone-melted commercial titanium alloys 77 Estimation of the fraction solidified at the initiation of beta-flecks 79 Estimation of distribution coefficients and fraction solidified at the initiation of beta-flecks with "pseudo-binary phase approach" 80 6.3 FORMATION MECHANISM OF BETA-FLECKS 82 Possibility of downward flow of liquid metal during solidification 82 Problems in the proposed formation mechanism of beta-flecks 84 7. CONCLUSIONS AND FUTURE WORKS 88 7.1 CONCLUSIONS 88 7.2 RECOMMENDATIONS FOR FUTURE WORKS 90 REFERENCES 92 APPENDIX A: MATHEMATICAL MODEL"METALS" 99 v

List of Tables Table 1 Distribution coefficients of alloying elements in titanium alloys 7,20-23 7 Table 2 Liquidus and solidus temperatures in titanium alloys!3,23,24 g Table 3 Melting experiment results with an argon arc furnace. 43 Table 4 Composition variations and distribution coefficients (k) in a zone melted 10-2-3 ingot. 53 Table 5 Composition variations and distribution coefficients (k) in a zone melted Ti-17 ingot. 53 Table 6 Composition variations and distribution coefficients (k) in a zone melted 6242 ingot. 53 Table 7 Composition variations and distribution coefficients (k) for oxygen and nitrogen in a zone melted 10-2-3 ingot 47. 53 Table 8 Parameters and values used for pseudo-binary phase approach for Ti-17. 81 Table 9 Parameters and values used for pseudo-binary phase approach for 10-2-3. 81 vi

List of Figures Figure 1 Solidification maps for (a) Ti-6-4 and (b) Ti-17 alloys 14. 24 Figure 2 Dendrite arm spacing in Ti-17 13. 24 Figure 3 Residual segregation index vs. homogenization parameter for chromium in steel 19. 25 Figure 4 Macrostructure of cross-section of a 10-2-3 ingot 27. 25 Figure 5 Optical micrograph and scanning fractograph of fractured 10-2-3 5. 26 Figure 6 Effect of beta-fleck area on LCFfifeof 10-2-3 5. 26 Figure 7 Typical microstructure of Ti-6-6-2 with beta flecks 28. 27 Figure 8 Various appearances of freckles in industrial castings 30. 28-29 Figure 9 Schematic diagram of directional solidification and associated thermal (pt), solutal (pc) and thermosolutal (pt+c) density profiles illustrating the density inversion theory 30. 29 Figure 10 Schematic illustration depicting freckle formation and associated fluid flow pattern 29. 30 Figure 11 The mechanism of freckle formation showing the sequence of the density-driven downward-forming channel to form a freckle 36. 30 Figure 12 Schematic illustration of typical curved growth front found in ingot 29. 31 Figure 13 Modified Rayleigh number vs growth front angle for alloy (a)cmsx-hb, (b)rene88, (c)nim80a, (d)in718-si, (e)waspaloy and (f)mar-m247 29. 32 Figure 14 Calculated Rayleigh numbers for the directionally solidification experiments for the SX-1 superalloy as a function of the thermal parameter G- 1/2»R- 1/4 42. 33 Figure 15 Photograph of an argon arc melting furnace in AMPEL. 41 Figure 16 Schematic diagram of an argon arc melting furnace in AMPEL. 42 Figure 17 Macrographs of 10-2-3 ingots melted by an argon arc furnace. (a)no.l (b)no.2 (c) No.3 (d) No.5 (e) No.7 56,57 Figure 18 Distribution of iron concentration in the horizontal direction in a 10-2-3 lboratory-melted ingot. 58 Figure 19 Distribution of iron concentration in the direction inclined to the horizontal direction by 30 in a 10-2-3 laboratory-melted ingot. 58 Figure 20 Distribution of iron concentration in the direction inclined to the horizontal direction by 45 in a 10-2-3 laboratory-melted ingot. 59 Figure 21 Distribution of iron concentration in the direction inclined to the horizontal direction by 60 in a 10-2-3 laboratory-melted ingot. 59 Figure 22 Evolution of temperature with time during solidification in Commercially Pure Titanium melted in an argon arc furnace. 60 Figure 23 Macrograph of a Ti-17 production ingot (as-received). 61 Figure 24 Microstructure of a Ti-17 production ingot. 61 Figure 25 Distribution of chromium concentration in a Ti-17 production ingot. 62 Figure 26 Estimated distribution of chromium concentration in a Ti-17 production ingot just after solidification. 62 Figure 27 Macrograph of a 10-2-3 production ingot (as-received). 63 Figure 28 Distribution of iron concentration in the direction inclined to the horizontal direction by 60 in a 10-2-3 production ingot. 63 Figure 29 Distribution of iron concentration in a 10-2-3 production ingot. 64 Figure 30 Macrographs of zone melted samples (as received). 65 Figure 31 Cross-sectional macrographs of a zone melted 10-2-3 ingot. 66 vii

Figure 32 Cross-sectional macrographs of a zone melted Ti-17 ingot. 67 Figure 33 Cross-sectional macrographs of a zone melted 6242 ingot. 68 Figure 34 Concentration distribution of alloying elements in the longitudinal direction of a zone melted 10-2-3 ingot. 69 Figure 35 Concentration distribution of alloying elements in the longitudinal direction of a zone melted Ti-17 ingot. 69 Figure 36 Concentration distribution of alloying elements in the longitudinal direction of a zone melted 6242 ingot. 70 Figure 37 Concentration distribution of alloying elements in the longitudinal direction of a zone melted 10-2-3 ingot (original data). 70 Figure 38 Distribution of oxygen concentration in the longitudinal direction of a zone melted 10-2-3 ingot 47. 71 Figure 39 Distribution of nitrogen concentration in the longitudinal direction of a zone melted 10-2-3 ingot 47. 71 Figure 40 Effect of iron concentration on density of the liquid and solid phase at around melting temperature (1905 K) in the 10-2-3 alloy. 72 Figure 41 Effect of temperature on density of the liquid and solid phase in the 10-2-3 alloy. 72 Figure 42 Effect of chromium concentration on density of the liquid and solid phase at around melting temperature (1914 K) in the Ti-17 alloy. 73 Figure 43 Effect of temperature on density of the liquid and solid phase in the Ti-17 alloy. 73 Figure 44 Effect of primary dendrite arm spacing on the Rayleigh number in a Ti-17 production ingot. 87 Figure 45 Effect of primary dendrite arm spacing on the Rayleigh number in a 10-2-3 production ingot. 87 viii

List of Symbols Symbols C Solute Concentration (wt.%) Co Equilibrium Solute Concentration (wt.%) CM Maximum Solute Concentration of Component i at Time th(wt.%) Cm Minimum Solute Concentration of Component i at Time th(wt.%) C M Maximum Initial Solute Concentration of Component i (wt.%) C m Minimum Initial Solute Concentration of Component i (wt.%) Cave Average Concentration of Alloying Element in the Matrix (wt.%) Cmax Maximum Concentration of Alloying Element Detected by EDX (wt.%) Cmin Minimum Concentration of Alloying Element Detected by EDX (wt.%) CL1,CL2 Liquid Composition in a Phase Diagram (wt.%) D s Interdiffusion Coeficient in the Solid State (m 2 /s) DT Thermal Diffusivity (m 2 /s) f s Fraction Solidified g Gravitational Acceleration (m/s 2 ) G Thermal Gradient (K/m) GF V Vertical Temperature Gradient (K/m) GL Temperature Gradient of the Liquidus (K/m) h Characteristic Linear Dimension (m) hm Height of the Mushy Zone (m) k Distribution Coefficient keq Equilibrium Distribution Coefficient K Permeability in the Vertical Direction (m 2 ) Km Mean Permeability (m 2 ) K y Permeability Parallel to the Primary Dendrite Trunks (m 2 ) L Half of the Dendrite Arm Spacing (m) Nf Cycle Number R Solidification Rate (K/s) Ra, RaT/s Rayleigh Number Racrit Critical Rayleigh Number Ra* Modified Rayleigh Number t Time (s) t s Total Solidification Time (s) T Temperature (K) Tl,T2 Temperature in a Phase Diagram (K) Tbuik Bulk Alloy Transformation Temperature (K) TL Liquidus temperature (K) Ts Solidus Temperature (K) V Withdrawal Rate (m/s) ix

Greek Symbols a Thermal Diffusivity of the Melt (m 2 /s) Si Residual Segregation Index Inclination Angle (degree) Dynamic Viscosity of Liquid (kg/m/s) Xi Primary Dendrite Arm Spacing (m) Secondary Dendrite Arm Spacing (m) Up Columnar Growth Coefficient for Primary Dendrite Arm Spacing ( m 1.25. s 0.25.K-0.5) Columnar Growth Coefficient for Secondary Dendrite Arm Spacing ( m>s 0.33) P Density (kg/m 3 ) po Reference Density (kg/m 3 ) pc Solutal Density (kg/m 3 ) PT Thermal Density (kg/m 3 ) pc+t Thermosolutal Density (kg/m 3 ) Pmatrix Density of Matrix (kg/m 3 ) pfreckle Density of Freckle (kg/m 3 ) Ap Density Difference (kg/m 3 ) Abbreviations AMPEL Advanced Materials and Process Engineering Laboratory EBR Electron Beam Remelting EDX Energy Dispersion Spectrometer EPMA Electron Probe Micro-Analysis ESR Electro-Slag Remelting HCF High Cycle Fatigue HDI High Density Inclusion LCF Low Cycle Fatigue LDI Low Density Inclusion N/A Not Applicable NIR Near-Infra Red PDAS Primary Dendrite Arm Spacing SDAS Secondary Dendrite Arm Spacing SEM Scanning Electron Microscope SX Single Crystal TC Thermocouple VAR Vacuum Arc Remelting UBC University of British Columbia

Acknowledgement I would like to thank first and foremost my supervisor, Dr. Alec Mitchell, for his invaluable guidance throughout this Master thesis. I would also like to thank my cosupervisor, Dr. Steve L Cockcroft, for his useful suggestions. I appreciate Mr. Rudy Cardeno and Ms. Mary Mager for their direct support through my experimental works. All the support staff in the- department of Metals & Materials Engineering at the University of British Columbia (UBC, Vancouver, Canada) were also helpful. I am very grateful for Timet Corporation and RMI Titanium Company to supply the materials and for the Wright Patterson Air Force Laboratory to conduct floating zone melting experiments. I really appreciate our company, Nippon Steel Corporation (Japan), for their support for everything I needed to make a living in Canada and to study at UBC. Finally, I would like to thank my wife who supports me every time and my family in Japan who encourages me. xi

1. INTRODUCTION Titanium and its alloys are promising structural materials for industrial use because of their high strength, low density and superior corrosion resistance 1-2. Since the price of titanium is higher than that of other conventional metals, such as steels, the titanium market is still small. However, the demand for titanium is projected to increase when titanium products are more generally recognized to have much longer life than other conventional metals since this leads to lifetime cost reduction. Titanium demand will also increase as the production costs are lowered. In titanium alloys, melting and casting problems are important issues 3, which can cause an increase in the production cost and also lower the quality of the products. The formation of Low Density Inclusions (LDI's), High Density Inclusions (HDI's) and betaflecks, caused by inhomogeneities in the cast ingot, have detrimental effects on the mechanical performances of titanium and its alloys 4 " 7. LDI's and HDI's are exogenous inclusions. The former originates from low quality raw materials, which contain locally high concentrations of nitrogen, oxygen or carbon, while the latter consists of heavy metal elements, like tantalum, cobalt or tungsten, from machine tool tips or heavy metal scrap 3. It has been shown to date that these exogenous inclusions can be reduced by the strict selection of raw materials or by applying hearth or skull melting processes 3-8 - 9. In contrast, beta-flecks, indigenous inclusions, are defined as localized areas rich in beta stabilizing elements 3. Beta-flecks are known to form through the segregation of beta stabilizing elements during solidification and have deleterious effects on mechanical properties, particularly on the fatigue life of titanium alloy components 5 " 7. Homogenization heat treatment, which results in the redistribution of beta-stabilizing

elements in products, is lengthy and causes an oxidizing problem on the surface of the products 3-10. The formation of beta-flecks during solidification or casting should be suppressed; however, the formation mechanism of beta-fleck has not yet been clarified and no effective manufacturing procedure for its elimination has been established. The present research study is focused on the segregation behavior of beta stabilizing elements in titanium alloys in order to clarify the formation mechanism of beta flecks. Chapter 2 contains the literature review on related topics. It summarizes segregation parameters of titanium alloys and describes the main features of beta flecks. The formation mechanism of freckles (melt-related defects in superalloys, arising due to fluid flow of the liquid metal in interdendritic region) is also discussed. The goal of the research study is presented in Chapter 3. Experimental procedures, conducted in the research study, are shown in Chapter 4. Results and discussions are presented in Chapter 5 and Chapter 6, respectively, followed by conclusions with recommendations for future works presented in the final chapter. 2

2. LITERATURE REVIEW 2.1 SEGREGATION BEHAVIOR OF ALLOYING ELEMENTS IN TITANIUM ALLOYS Solidification microstructures in titanium alloys Prior to a discussion of beta-flecks in titanium alloys, it is important to review segregation behavior in titanium alloys. In the production of large VAR (Vacuum Arc Remelting) ingots, the solidifying interface is in a cellular mode in CP (Commercially Pure) titanium and in a cellular or dendritic mode in the 6-4 (Ti-6%A1-4%V) alloy 7 ' 11-49. In fact, both dendritic 14 and non-dendritic 11-49 microstructures have been reported for the solidification microstructure of 6-4 VAR ingots, which might be because the chemical composition of 6-4 corresponds to a transition area from a cellular to a dendritic solidification manner. In contrast, solidification proceeds in a columnar/equiaxed dendritic mode in beta alloy VAR ingots 7 ' 11-49. This assumption has been clarified recently by observing the surface of the solidification interface in CP and beta alloys 11 and the dendritic morphology of surfaces of shrinkage cavities 12-13 in beta alloys. A wider temperature range between the liquidus and the solidus line in beta alloys than in CP and 6-4 is considered as a reason for a change in the solidification manner. Therefore, microscopic segregation of alloying elements has to be mainly taken into account in beta alloys, while macroscopic segregation might be recognized in CP and the 6-4 alloy in most cases. Nastac et al 14 proposed solidification maps for the 6-4 alloy and Ti-17(Ti-5%Al-2%Sn- 2%Zr-4%Mo-4%Cr) by using a software modified from SIMCAST, as shown in Figure 1. According to the results, typical cooling conditions of a production ingot (Calculated with 3

VAR model (Ingot Dia.=762 mm, Ingot Length=635 mm, Melting Rate=273 kg/hr): G=5-10 K/cm=5xl0 2 -lxl0 3 K/m, V=40 um/s=4xl0 5 m/s, R=G»V=2-4xlO" 2 K/s) produce a mixed microstructure of columnar/equiaxed grains. Solidification maps are very useful in predicting microstructures under certain solidification conditions and some maps have been developed for stainless steels 15. It is important to produce solidification maps for titanium alloys, although the practical problems of experimental verification are considerable. Dendrite arm spacing Dendrite arm spacing is an important morphological parameter when attempting to model beta-fleck formation 11-13. However, no effective solution has been developed for etching titanium alloys, which offers direct and clear observation of dendrite structures 11. This is due to the transformation from beta to alpha phase that occurs during cooling, which makes it difficult to identify the prior dendrite structures. Dendrite arm spacing has been evaluated indirectly from the spacing between concentration peaks of alloying elements obtained from Electron Probe Micro-Analysis (EPMA), however; only a few results have been reported 16-17. The relationship between dendrite arm spacing and cooling conditions in binary titanium alloy ingots is shown in Figure 2 13-16. Primary dendrite arm spacing (PDAS) and secondary dendrite arm spacing (SDAS) are assumed to be between 2500-4000 um and 1500-2000 um, respectively, at R=G»V=2-4xl0" 2 K/s, which is a typical cooling condition for production ingots. Ichihashi et al 49 demonstrated 800-1000 um for SDAS in 660mm dia. beta alloy ingots. Aleksandrov et al 17 reported 3000-5000 urn for PDAS in 430mm dia. BT3-l(Ti-6.5%Al-2.5%Mo-1.5%Cr-0.5%Fe-0.3%Si) production ingots. 4

Dendrite arm spacing was calculated on the basis of the following equations by Nastac 14 and the results are shown in his solidification maps (See Figure 1). Xi = u P «V-0-25.G-«-5 (Eq.l) X2 = Us»t s - -33 (Eq.2) where Xi is the PDAS(m), \x P is the columnar growth coefficient(=1.924xl0-3 m 1-25^0-25 *!^- 0-5 in Ti-17), V is the withdrawal rate(m/s), G is the thermal gradient(k/m), is the SDAS(m), \x s is the columnar growth coefficient^lgssxlo- 5 m«s 0-33 in Ti-17) and t s is the local solidification time(s) The above equations are formulated on a basic theory that the primary arm spacing depends on the cooling rate during solidification and the secondary arm spacing is controlled by the local solidification time 15. Equations 1 and 2 yield 1000-2000 jam for PDAS and 200-400 (a,m for SDAS for a Ti-17 production ingot. These values are smaller compared to the experimental values and the difference between them is presumed to be caused by the limited number of data on dendrite arm spacing in titanium alloys. In this case, experimental values are considered to be more reliable, although the number of the data is limited. A comparison of the dendrite arm spacings in titanium alloys with those in superalloys and steels was also attempted. McLean's morphology map of superalloy structures 18, which was established by a large number of data, can be used as a guideline. From the morphology map, at V»G=2-4xlO- 2 K/s, SDAS is estimated as 180-240 j.m, which is smaller than that reported on titanium alloys. Under similar solidification conditions, PDAS and SDAS were 1000-1500 nm and 200-400 \im in Fe-26%Ni alloy 19, which are also smaller than those observed in titanium alloys. In conclusion, under conditions for casting large 5

production ingots, the dendrite arm spacing is smaller in titanium alloys than in superalloys or steels. Solidification parameters The solidification parameters related to the segregation behavior of alloying elements in titanium alloys, k(distribution coefficient), TiXliquidus temperature) and Ts(solidus temperature) are critical. In particular, the distribution coefficient of the alloying element is essential, since the volume fraction of the solid phase in the liquid/solid interface can be estimated by applying the distribution coefficient to the Scheil equation, shown as the following (Eq.3). C s =k.co»(l-f s ) k - 1 - (Eq.3) where C s is the concentration of solute in the solid, Co is the average concentration of solute in the solid, k is the distribution coefficient and f s is the fraction solidified Distribution Coefficients Distribution coefficients or partition coefficients, k, have been reported in some studies 7 ' 20 " 23. Under equilibrium conditions, k is the ratio of a slope of liquidus line to that of the solidus line in binary phase diagrams at a given composition 20. Tin, zirconium, vanadium, chromium and iron have k<1.0, while aluminum, molybdenum, oxygen, nitrogen and carbon show k>1.0. This indicates that the former five elements tend to be depleted in the initial part of solidification and to be enriched in the final part, while the latter five behave in the opposite way. Distribution coefficients obtained from experiments 7-21 ' 22 and calculation 23 are listed in Table 1. Different alloys were used in each reference (See Table 1). In all the studies, the k value for tin deviated from those 6

obtained from the binary phase diagram, whereas in ref(7), the k values for chromium and iron all deviated from those obtained from the binary phase diagram. In contrast, aluminum and vanadium show distribution coefficients close to those obtained from the equilibrium phase diagrams. Tin is neutral in stabilizing phases in titanium alloys, which might be the cause for the variation in the distribution coefficients. However, the cause of variation in iron and chromium in ref(7) from equilibrium data is not clear. These results suggest that in many cases the binary equilibrium phase diagram can be used to obtain practical phase conditions but there are some exceptions like iron, chromium, etc. Therefore, it is important to measure distribution coefficients of alloying elements in commercial titanium alloys with multi-component system. Table 1 Distribution coefficients of alloying elements in titanium alloys 7-20 - 23 Al Sn Zr Mo V Cr Fe O N C note 1.05 0.92 0.90 1.50 0.95 0.70 0.60 1.60 1.58 0.50 Binary phase diagram: ref(20) 1.01-1.05 1.06-1.51 0.95-1.01 0.87-1.03 BT3-1: ref(7) 1.00-1.06 1.03-1.14 0.77-0.84 1.14-1.18 0.90 0.77 0.59 679, 6-6-2, Ti-Mo-Cr: ref(21) 1.02-1.08 1.09-1.13 0.89-0.95 0.79 0.61-0.71 6-4, 6-6-2, 15-3 : ref(22) 1.05 0.83 0.92 1.06 0.65 Ti-17: ref(23) BT3-1: Ti-6.5Al-2.5Mo-l.5Cr-0.5Fe-0.3Si, 679 : Ti-2.25Al-llSn-5Zr-lMo-0.22Si 6-6-2 : Ti-6Al-6V-2Sn, 6-4 : Ti-6A1-4V, 15-3 : Ti-15V-3Al-3Cr-3Sn Ti-17 : Ti-5Al-2Sn-2Zr-4Mo-4Cr 7

Solidus and Liquidus Temperature Auburtin 13 presented experimental data on Ti-6242(Ti-6%Al-2%Sn-4 0 /ozr-2%mo), Ti-17 and 10-2-3(Ti-10%V-2%Fe-3%Al) obtained from thermo-couple measurements (TC data) and from pyrometer measurements (NIR data), as shown in Table 2. Table 2 includes the calculation results on Ti-17 reported by Nastac 23, including values obtained from Metals Handbook 24. The liquidus temperatures obtained as NIR data are lower than the TC data. The measurements by pyrometer are subject to calibration errors in emissivity values and temperatures obtained from thermocouple measurements are possibly more accurate. TC data on the liquidus and solidus temperatures listed in the table, therefore, are considered to be more reliable. Table 2 Liquidus and solidus temperatures for titanium alloys 1323-24 Alloy T liq ( C) T so i( C) notes Ti-6242 166815 1634110 TCdata ref(13) Ti-6242 1675 1595 Metals Handbook data ref(24) Ti-6242 1590 N/A NIR data ref(13) Ti-17 164115 1632110 TCdata ref(13) Ti-17 1590 N/A NIR data ref(13) Ti-17 1649.3 1591.1 PAM ingot model ref(23) Ti-17 1649.3 1503.1 Scheil model ref(23) Ti-17 1649.3 1625.0 Lever rule model ref(23) Ti-10-2-3 163215 1595110 TCdata ref(13) Solid State Diffusion When partition or segregation coefficients are considered, the diffusion behavior of solute atoms immediately after solidification has to be considered as the back diffusion effect of the solute atoms cannot be ignored 19-25 - 26. Shamblen 10 calculated the interdiffusion coefficient (Ds) of chromium in the solid state of Ti-17 on the basis of experimental data. This is shown as (Eq.4), which was estimated from the relationship 8

between diffusion coefficients and temperatures in the literature. D s = -1.93xl0-4 x(l/t( C))+1.55xl0-7 (cm 2^- 1 ) (Eq.4) where D s is the interdiffusion coefficient of chromium in Ti-17(cm 2 «s- 1 ) and T is the temperature( C). It is possible to estimate the concentration of chromium just after solidification in Ti-17 from the relationship between residual segregation index(8i), shown as (Eq.5) and D s»th/l 2, as presented in Figure 3 19, where D s is the interdiffusion coefficient of solid state (m 2 «s- 1 ), th is the local solidification time (s) and L is the half of the dendrite arm spacing (m). 5i = (CM-Cm)/(Co m -Co M ) (Eq.5) where 5i is the residual segregation index, CM is the maximum solute concentration of component i at time th (s), Cm is the minimum solute concentration of component i at time th (s), C m is the maximum initial concentration of component i (wt.%) and C M is the minimum initial concentration of component i (wt.%). The above equations were applied for estimating the redistribution of segregated alloying element during the homogenizing process which consists of 1 to 100 hours of heat treatment. Moreover, it is clear from (Eq.4) that slow diffusion rates make homogenization of defects quite difficult. Holding time in the homogenizing process is considered to be much longer than the corresponding holding time during solidification but the effect of solute redistribution during solidification should be taken into consideration. 9

2.2 BETA-FLECKS IN TITANIUM ALLOYS Features of beta-flecks Beta-flecks are localized defects, which contain a higher content of beta stabilizing elements than in the bulk. Beta-flecks have been found frequently in BT3-1, BT-22 (Ti- 5%Al-4.5%Mo-4.5%V-l%Cr-l%Fe), Ti-17, 10-2-3 and other auoys: au of which contain iron and/or chromium. Iron and chromium are strongly rejected at the solid/liquid interface with effective segregation coefficients of 0.6-0.8 in most titanium alloys 13, causing macrosegregation and microsegregation. The former occurs under a high temperature gradient and the latter takes place when the interface is dendritic 3. Therefore, microsegregation should be considered in production scale ingots, while macrosegregation might take place under quite special solidification conditions, such as an uni-directional solidifying condition with a high thermal gradient and a slow cooling rate using a floating zone melting furnace. Since there has been no clear criteria established for the amount of segregation responsible for beta-flecks, the chemical compositions of areas corresponding to beta-flecks are reported to be different depending on the observer. In 10-2-3 ingots, Brooks found beta-flecks formed with an enrichment of 0.4wt.% iron and lwt.% vanadium 27. Zhou found that beta-flecks contained at least 2.72wt.% iron and 10.6wt.% vanadium 6. The largest difference in iron content observed was 0.5% by Chen 4. Shamblen 10 demonstrated that a typical beta-fleck might show increases of 1.0-1.5wt.% chromium and 0.5wt.% zirconium along with decreases of 0.5wt.% molybdenum and 0.2wt.% aluminum in Ti-17. Tetyukin 7 presented increases of 0.25wt.% chromium and 0.lwt.% iron and a decrease of 0.5wt.% molybdenum were found in "strings"(beta-flecks) in BT3-1. In fact, these results were obtained from chemical analysis conducted on the locations showing beta-fleck 10

microstructures, which were identified by each observer. Observed microstructures, including beta-flecks, might have appeared differently even if the chemical compositions were the same, depending on the etching procedures or techniques. This might have caused different chemical compositions of beta-flecks, depending on the observers. The appropriate definition for beta-flecks has to be established to identify beta-flecks. With respect to heat treatment/forging practice, practical criteria to define the chemical compositions of beta-flecks have been proposed 13. In heat treatment/forging operations, problems arise from beta-flecks due to differences in the transformation temperature between beta-flecks and the matrix, caused by a difference in chemical composition. A heat treatment just below/above transformation temperature is often required in titanium alloys, where a difference in the transformation temperature in the beta-fleck and in the matrix can give rise to a decrease in mechanical properties. Therefore, most product specifications for titanium alloys specify a permissible range of heat treatment temperature relative to the bulk alloy transformation temperature, e.g. (Tbuik-15 C) and the industrial experience of the occurrence of beta-flecks is related to this interval in practice 13. For example, in BT3, the bulk composition contains a chromium level of 1.5 wt.%, and a typical beta-fleck area will contain 1.7-1.8 wt.% chromium. In 10-2-3, the bulk composition is 2.0 wt.% iron and the beta-fleck area will contain 3.1 wt.% iron. In Ti-17, the bulk is 4 wt.% chromium and the beta-fleck will contain 5.5 wt.% chromium 3. This definition is quite reasonable and can easily be applied to identify betaflecks. From these chemical compositions, the fraction solidified can be estimated at the point when beta-flecks are formed. For instance, if these alloys are assumed to solidify under conditions obeying the Scheil equation, with distribution coefficients of 0.60 for iron and 11

0.70 for chromium, temperatures at which beta-fiecks initiated are calculated to be consistent with the fraction solidified, is, of 0.91 for iron and 0.89 for chromium. This high fraction solidified suggests that beta-flecks were initiated at the final stage of solidification. Morphologically, beta-flecks observed in large ingots appear as string or pencil like structures and V-shaped distributions near the center of the ingots 7-27. Tetyukin 7 demonstrated bright and string-shaped streaks in the equiaxed/columnar dendritic region around the center of a 750 mm dia. BT3-1 ingot by radiographic observation. This result suggested that the density of the beta-fleck might be quite different from that of the bulk, although the detailed conditions of the radiographic method were not mentioned. Brooks 27 reported V-shape distributions of beta-flecks both with iron distribution mapping analyzed by X-ray spectroscopy and with optical microscopy. It was found that beta-flecks, in which iron was enriched compared with the matrix, corresponded to the termination of magnetic stirring during the melting process and that the V-shape represented the bottom shape of molten pool. A clear microstructure of beta-flecks in an as-cast ingot was obtained, which appears as dark-colored pencil-like contrasts, as indicated in Figure 4, after a "beta-fleck heat treatment" (800 C for lhour and water quenched), projecting to obtain a clearer microstructure of beta-flecks. The noticeable feature in this figure is that beta-flecks run through the beta grains. This result suggests that the beta grains might have recrystallized irrespective of segregation of beta stabilizing elements during the "betafleck" heat treatment. Also, it has been reported that a beta-fleck consists of a prior-beta grain or some grains and the grain size was larger than that in the matrix in forged and heat-treated products, as shown in Figure 5 5 During the solution heat treatment, beta grains must have 12

recrystallized and the grain boundaries should have formed just at the interface between the original beta-flecks and the prior matrix. Larger grains in the beta-fleck area might have formed because of a higher growth rate in the recrystallized grains due to a lower transformation temperature in the beta-fleck area than that of the matrix. However, the reasons for this behavior are not clear. Further detailed studies on the microstructural evolution of beta-flecks during and after heat treatment are necessary to clarify the effect of beta-flecks on mechanical performances, which will be discussed below. Effects of beta-flecks on mechanical properties The effects of beta-flecks on mechanical properties of titanium alloy products have been studied and in most cases the detrimental effects have been demonstrated 4-7 - 28, particularly, on Low Cycle Fatigue (LCF) life 5-7 - 28 and fracture toughness 4-5. All of the authors used forged and heat-treated final products, which means that beta-flecks survived even after the forging process to reducefinalmechanical properties. Chen 4 found that yield strength increased from 1176 to 1306 MPa and the total elongation decreased from 6 to 1 % due to the existence of beta-flecks in 10-2-3. Since Kevex microprobe analysis revealed a high peak concentration of iron in beta-fleck area, solid solution hardening should have occurred in the area. However, no detailed discussion on microstructures was given. Zhou 5 reported a detrimental effect of betaflecks on LCF life, considering both the volume fraction and the maximum size of betafleck in 10-2-3, as shown in Figure 6. A decrease in LCF life could be found in specimens containing even a small volume fraction of beta-flecks, which did not affect the tensile properties in the whole specimen. In the specimens, cracks were initiated from the betafleck, leading to a decrease in LCF life. In general, a particle, which is harder than the 13

matrix, could be a stress raiser and may cause cracks at the interface between the particle and the matrix due to a stress concentration 57. Thisfindingsupports Tetyukin's results on BT3-1 6, in which LCF life decreased in the region containing beta-flecks, although the tensile strength of the region was almost the same as that of beta-fleck free region. According to detailed observations of microstructures, Funkenbusch 7 demonstrated that beta-flecks were crack initiation sites in Ti-17, which caused a shorter LCF life. In contrast, Rudringer 28 showed that beta-flecks did not affect the LCF and High Cycle Fatigue (HCF) life in Ti-6-6-2 (Ti-6%Al-6%V-2%Sn). Actually, no clear difference could be found in LCF and HCF life between samples containing beta-flecks and samples free from beta-flecks. This might be because the author used 6-6-2, which contains vanadium as a beta-stabilizing element. As shown in Table 2, the distribution coefficient of vanadium is 0.9-0.95 and the segregation ratio of vanadium in 6-6-2 is assumed to be much lower than that of iron in 10-2-3 or that of chromium in Ti-17. The degree of solid solution hardening might be much less in 6-6-2 than in 10-2-3 or Ti-17. However, microstructures appeared differently near the center area compared with those in other locations in 6-6-2 ingots, which might have been affected by beta-flecks, as presented in Figure 7 28. This result shows that there are two types of beta-flecks: one being harmful and the other being harmless to mechanical properties. A difference in alloy type of the bulk materials could have caused different effects in mechanical properties of titanium alloys containing beta-flecks: 6-6-2 is an alpha-beta alloy, while 10-2-3 and Ti-17 are beta-alloys. Solid solution hardening in the beta phase might have more deleterious effect on mechanical properties in beta alloys than in alphabeta alloys. In this case, however, the effect of alloy type must be lower than that of alloying elements because LCF life was reduced by beta-flecks in BT3-1, which is an 14

alpha-beta alloy containing chromium and iron. In conclusion, beta-flecks in general tend to reduce the mechanical properties in titanium alloys which contain chromium and/or iron. Beta-flecks cause a decrease in tensile elongation if the volumefractionis high or a decrease in LCF life if the size is considerably large. However, detailed analysis has not yet been carried out and the quantitative effects, such as volume fraction, size and hardness of beta-flecks. Formation of beta-flecks is one of the most important issues for titanium alloys. Detailed and systematic research work is considered to be necessary. Features of freckles Freckles (channel segregates or 'A'-segregates) are melt-related defects in nickel-based superalloy or specialty steel castings, which appear as a long trails of the equiaxed grains with a composition shift consistent with alloy segregation 29-30. Freckles are also highly undesirable in critical applications because of their deleterious effect on mechanical performances. Because of their obvious similarity to beta-flecks, it is worth while to review the literature on freckle formation. Features of freckles, together with comparison with those of beta-flecks, are given below: (1) Freckles show various appearances depending on a difference in solidification procedure and alloying system. (See Figure 8 30 ) For example, in VAR/ESR(Electro-Slag Remelted) ingots, freckles are usually located in the center to mid-radius of the billet 31 and in directionally solidified superalloy castings (DS or SX), freckle lines are normally located on the exterior surface of the castings 32. In killed steel ingots, freckles ('A'-segregates) usually form in the middle of the solidification zone, which grows perpendicularly to the sidewalls 33 and in IN718 15

containing small amounts of silicon (Ni-0.5%Al-0.2%Co-18%Cr-3%Mo-5%Nb-55%Nil%Ti-Fe, Si<0.3%), the freckles distributed parallel to the liquidus line 30. (2) Freckles are found to be enriched in the normally segregating elements and depleted of the inversely elements. (3) The freckle initiation temperature is assumed to be consistent with a fraction solidified fs=0.5 29. (4) Freckling can be significantly reduced and even avoided by operating at larger thermal gradients and faster solidification rates 30. The location and the shape of freckles is influenced by the mushy zone and its shape 34, which indicates appearance of freckles changes depending on the casting procedures and alloying systems. Geometric distribution of beta-flecks is in most cases V-shape in the center of ingots 7-27, which appears similar to that observed in IN718 containing low Si. The fact that freckles are enriched in normally segregating elements indicates that freckles are shifted toward the eutectic composition 35. This behavior supports the fact that superalloys, which contain high titanium (segregating normally) or tungsten (segregating inversely), are reported to be more freckle prone 35. The frecle initiation temperature is consistent with f s =0.5, which is quite different from f s =0.8-0.9 for beta-flecks 13. It would appear that the geometric distribution and the initiation temperature are quite different in freckles and in beta-flecks. 16

Formation mechanisms of freckles and beta-flecks, It is now generally agreed that freckles arise due to channels associated with "thermosolutal" or "double diffusive" convection in the mushy zone, which is caused by a density inversion in the mushy zone, as shown in Figure 9 30. In this figure, the alloy is solidifying vertically upward, while the heat flow is vertically downward, creating a vertical thermal gradient along the casting. In addition to a thermal gradient, there also exists a variable solute concentration gradient in the liquid between the bottom of the mushy zone and the top of the casting. The density of a liquid alloy is dependent on its temperature and solute concentration, whose profiles are also indicated in the figure (in this case, rejected solute is lighter than the solvent). Given such a density profile, it can be seen that the interdendritic liquid lower in the mushy zone (enriched in solute) is less dense than the liquid at the dendrite tip. This is a case of density inversion at the growth front. This system is unstable, and can lead to fluid convection in order to reduce the potential energy. This phenomenon is known as "thermosolutal" or "double diffusive" convection and is considered to be the cause of freckling and the formation mechanism of freckles is schematically shown in Figure 10 29. The rising plume has a steady-state lifetime during which it collects interdendritic liquid by fluid movement in a direction approximately at right angles to the growth direction and is established over one or more primary dendrite spacings. The freckle channels eventually freeze, as the thermal profile passes through the region. This mechanism can explain freckles formed in the direction parallel to the growth direction. The segregation channels corresponding to an array of freckles were formed parallel to the liquidus line in IN718 containing low silicon, in which density inversion during solidification was not obtained, based on calculations by Auburtin 35. The author also 17

computed interdendritic liquid density profiles by "METALS" and clarified that freckles in this alloy are heavier than the surrounding.bulk metal (pmatrix= 7490 kg/m 3 (Tn q =1336 C), pfreckie= 7570 kg/m 3 (Tu q =1336 C), pfreckie= 7640 kg/m 3 (T S oi= 1260 C)). VanDenAvyle 36 has schematically described this mechanism of freckle formation, as shown in Figure 11. When a liquid of composition CL2 is increased in temperature from Tl to T2, the composition of the liquid will tend to decrease to CL1 by remelting some of the surrounding solute-lean solid. This dissolution process, resulting from interdendritic liquid flowing into a higher temperature field, is the basis of the mechanism by which the channel defects form and propagate. From optical microscopic observation and microprobe analysis, it appeared that these defects formed fairly deep in the mushy zone 30. It is quite interesting that even in the same alloy system, in this case in nickel-base superalloys, completely different phenomena can occur depending on the relationships between the density of liquid and that of solid. It is therefore important to take the effect of the density into consideration during solidification. Brooks proposed a density-driven downward-forming channel for the formation mechanism of beta-fleck in large ingots, as shown in Figure 11, which was used for freckles in IN718 containing low silicon 27. A requirement for this type of channel to form is that the density of the interdendritic liquid increases during solidification process. This phenomenon is possible because chromium and iron tend to concentrate in the liquid and are heavier elements than titanium. Auburtin 13 calculated densities of beta-flecks and that of bulk liquid at 1605 C in 10-2-3 by "METALS" and obtained 4243 and 4184 kg/m 3, respectively. The data obtained by Zhou 5 were used as chemical compositions of betaflecks and bulk liquid(fe; 3.10 wt. /o(beta), 2.03 wt.%(bulk), Al; 2.25 wt.%(beta), 3.05 wt.% (bulk), V; 11.00 wt.%(beta), 10.23 wt.%(bulk)]. This calculated result is considered to 18

support Brooks' mechanism, although the density gradient is very small. Freckle Criterion and its application to beta-flecks The development of a numerical criterion, which provides quantitative insight on the conditions of freckle/beta-fleck formation, is considered as a major factor toward the successful manufacture of large diameter ingots. There have been some approaches, which have attempted to clarify the criteria of freckle formation recently. Auburtin 29 has first adopted the basic Rayleigh number suggested by Sarrazin and Hellawell 37. The basic Rayleigh number represents the ratio of driving force for flow to resistance against flow and may be employed to characterize the onset of fluid flow in unstable systems 38. He attempted to calculate the basic Rayleigh number by putting the growth front angle of freckles to the horizontal direction as a geometrical factor but could not obtain a single value for the Rayleigh number, above which no freckling occurs. In practice, the mushy zone is curved against the growth front angle in most cases and the direction of permeation relative to that of gravity should be considered. Permeability of liquid metal was considered in a typical mushy zone arising in VAR processed ingots, as shown in Figure 12. Considering the relationship between permeability and primary/secondary dendrite arm spacing geometrically 39 " 41 together with the relationship between dendrite arm spacing and temperature gradient, Auburtin finally obtained the modified Rayleigh number, Ra*. For each experimental casting, Ra* was plotted against growth front angle, as presented in Figure 13. The freckled and freckle-free regions were clearly divided by a horizontal line and a critical threshold Ra* value was achieved for each alloy. For example, Ra*(CMSX-llB)=0.88, Ra*(IN718-Si)=0.65, etc. Auburtin's approach is simple and the obtained threshold value is considered to be 19

accurate and reliable. However, experiments are necessary to determine the threshold value for every alloy system because the value might be different depending on alloy systems. Beckermann 42 proposed the critical Rayleigh number, which can be applied to any solidification conditions, any alloying systems and so on. Beckermann's approach adopted a characteristic linear dimension that is different from Auburtin et al 29. The proposed Rayleigh number has a maximum value for 10-15vol.% of fraction solidified. Once the maximum Rayleigh number is found, it needs to be compared to some critical value(ra C rit) to judge the stability to freckling. Here, the critical Rayleigh number is defined such that freckles will not form if Ra<Ra C rit (Eq.6) where Ra is the Rayleigh number, Racrit is the critical Rayleigh number. The relationship between Ra and G- 1/2 «R _1/4 is shown in Figure 14, which was originally obtained by Pollock 43. Beckermann proposed Racrit = 0.25 as a criterion for freckling, considering a more conservative value than G- 1/2»R- 1/4 < 0.95 proposed by Pollock, which corresponds to Racrit = 0.4. This critical value should be the same for all superalloys, assuming a minimal variation with other system parameters. Beckermann evaluated the critical Rayleigh number from numerical simulations as well, which predicted the possibility of channeling leading to freckle formation. He also observed the effect of inclination angle to the critical Rayleigh number and clarified the relationship as the following: Racrit = 0.125-0.00144 for 10 < <(><45 (Eq.7) where <() is the inclination angle (degrees). Beckermann's approach is also simple and universal to all the superalloy products, 20

irrespective of alloy systems and casting procedures. However, in some cases, the critical value might be too conservative, which restricts the production and processes too much. These two different approaches give us important information on criteria for freckling and they also have their own advantages and disadvantages. There has been no report on criteria for the formation of beta-flecks in titanium alloys yet. Only Auburtin 13 tried to estimate the dendrite arm spacing, which causes the densitydriven liquid flow in Ti-10-2-3, according to the basic Rayleigh number, Ra, shown as (Eq.8). The following parameters were used to calculate Ra, in which A,i,primary dendrite arm spacing (PDAS), was substituted for h. Ra-r/s = g«dp/dz/(r «D T /h4) (Eq.8) where g is the gravitational acceleration(m/s 2 ) (= 9.81 m/s 2 ), DT is the thermal diffusivity (m 2 /s) (= 9x10-6 m 2 /s), r\ is the dynamic viscousity of liquid titanium(kg/m/s) (= 0.004 kg/m/s), h is the characeristic linear dimension(m) and dp/dz is the density gradient (kg/m) (calculated from VAR model 58 ). When Ra<l, density driven fluid flow is unlikely; when Ra>l, density drivenfluidflow is more likely to occur 37. The above equation takes into account permeability due to dendrite arm spacing and is sensitive to X. Using a calculated liquid density gradient, the primary dendrite arm spacing should be between 1000 and 1500 um before density driven fluidflowwould occur, according to the Rayleigh criterion. 21

2.3 SUMMARY OF LITERATURE REVIEW Studies on segregation of beta stabilizing elements in titanium alloys during solidification were reviewed and can be summarized as : (1) In large scale ingots, the solidification mode in CP is either planar or cellular and that in 6-4(Ti-6%Al-4%V) is either cellular or dendritic, while solidification proceeds in a columnar/equiaxed dendritic mode in beta alloys. (2) Dendrite arm spacing in titanium alloys is in the same order of those reported on steels or copper alloys. Data on titanium alloys is limited and some of them might not be reliable. (3) Experimental distribution coefficients are close to those obtained from the equilibrium binary phase diagram. However, some alloying elements, like iron or chromium, have different distribution coefficients in different alloy systems. (4) By using the Scheil equation, temperatures at which beta-fleck initiated were estimated to be consistent with the fraction solidified of 0.8-0.9. (5) Beta-flecks in as-cast ingots appear irrespective of the etched microstructures, while those after heat treatment contain a prior-beta grain or some prior-beta grains. (6) Beta-flecks have detrimental effects on LCF (Low Cycle Fatigue) life in alloys containing chromium and/or iron. There have beenno detailed studies on relationship between mechanical properties and microstructures. (7) Differences in features of freckles and those of beta-flecks are : i) Geometrical distribution in large ingots or billets Freckles : along the longitudinal direction in the center to mid-radius of the billet Beta-flecks : V-shape distribution in the center of the ingot 22

However, in low Si-IN718, freckles are distributed parallel to the liquidus line, ii) The initiation temperature Freckles : f s (fraction solidified) = 0.5 Beta-flecks : f s = 0.8-0.9 (8) The density-driven upward "thermosolutal channel" is proposed as the formation mechanism of freckles in superalloys except for low-si IN718. In low-si IN718, the density-driven downward channeling is proposed for the freckle formation mechanism. The formation mechanism of beta-flecks is estimated to be similar to the latter. (8) For some superalloys, criteria for freckle formation are clarified on the basis of the modified Rayleigh number. For titanium, no criterion has been made clear for formation of beta-flecks yet. Only secondary dendrite spacing, which cause density drivenfluidflow,was estimated according to the basic Rayleigh number. 23

Range of VAR ingot values for Ti alloys at 10000 1000 i c 100 a. to c 0.001 0.01 0.1 10 100 1000 Cooling Rate (G x R) C/see Figure 2 Dendrite arm spacing in Ti-17 1 3. 24

Figure 3 Residual segregation index vs. homogenization parameter for chromium steel 19. (a) Longitudinal direction (b) Radial direction Figure 4 Macrostructure of cross-section of a 10-2-3 production ingot 27. 25

Figure 5 Optical micrograph and scanning fractograph of fractured 10-2-3 5 (Forged + 760Cx2hr WQ + 520Cx8hr AC) o Ti-10%V-2%Fe-3%AI alloy 760Cx2HR WQ + 520Cx8HR AC Low Cycle Fatigue Test Constant amplitude longitudinal pull-pull cyclic stress Stress Ratio : R-0.1 The Cyclic Frequency: f=15hz o maximum beta-fleck area volume fraction of beta-fleck beta-fleck area =31.22% 6 10 12 beta-fleck area (%) Figure 6 Effect of beta-fleck area on LCF life of 10-2-3 5. 26

Figure 7 Typical microstructures of Ti-6-6-2 with beta flecks 28. 27

concentrates under hot-top segregation bands A-segregates V- segregates cone of negative segregation a) "A" segregate in a large killed steel ingot b) Centre to mid-radius freckles in VAR IN718 (quarter of a cross-section) Figure 8 Various appearances of freckles in industrial castings 30 (continued). 28

c) Surface freckles in the root portion of a large SX IGT Mar-M247 blade Figure 8 Various appearances of freckles in industrial castings 30. Figure 9 Schematic diagram of directional solidification and associated thermal(p T ), solutal(pc) and thermosolutal(pt+c) density profiles illustrating the density inversion theory 30. 29

Liquid Melt FreckJe Plume A A Heavier Non- _Segregatecj Liquid Equiaxed I and/or I jeutectk-; j Enriched j I Material i (l-2m«i) f Lighter «Segregated ' uquicl Figure 10 Schematic illustration depicting freckle formation and associated fluid flow pattern 29. T, <;Y- X r / " '"\ Sy c L % Nb a) An alloy 718 niobium pseudobinary phase diagram. When liquid of C L 2 is increased from T 2 to Ti, C S r- Csi and C L2 -C L1, b) Increased density of interdendritic liquid results in a downward flow, c) Channel defects form by a d) The channel consists of a highdissolution mechanism, solute dendritic fragmented region. Figure 11 The mechanism of freckle formation showing the sequence of the densitydriven downward-forming channel to form a freckle 36. 30

31

2.5 E op 1.5 Freckles o No Freckles I Ra* [CMSX-11B \ 2.5 3 z 1.5 T Freckles o No Freckles Ra* (Rene88 1 T3 <U s ' 3 o 0.5 1 i V~ (No Freckles) I 0.5 [No Freckles] 10 20 30 Growth front angle (deg.) 40 10 20 30 Growth front angle (deg.) 40 (a) 2.5! Freckles i i o No Freckles j j Ra*! [Nim80A ) 5 1.5 Freckles! * No Freckles Ra* (ln718-si ) 1.5 Bi a (Freckles i 1 (Freckles ) o S 0.5 (No Freckles ) -g 0.5 (No Freckles ) 10 20 30 Growth front angle (deg.) (C) 40 10 20 30 Growth Front Angle (deg.) id) 40 e 3 Z 2.5 1.5 j. Freckles j j o No Freckles;! i Ra* (Waspaloy \ S 3 z 00 3.5 3 2.5 2 Freckles j o No Freckles: Ra* I Mar-M247 ) OS -a o 2 0.5 (No Freckles \ I 1.5 t 1 0.5 Freckles [No Freckles] 10 20 30 Growth Front Angle (deg.) (e) 40 0 10 20 30 Growth Front Angle (deg.) Figure 13 Modified Rayleigh number vs. growth front angle for alloy (a)cmsx-llb, (b)rene88, (c)ni80a, (d)in718-si, (e)waspaloy and (f)mar-m247 2 9. if) 40 32

100 10 <» No Grain Defects x Freckles and/or Grains s 1 Z 5 I 0.1 <2 o.oi 4 critical value proposed by Pollock and Murphy (1996) o.ooi o.i G-, / J R- w [cm, VV w ] 10 Figure 14 Calculated Rayleigh numbers for the directionally solidification experiments for the SX-1 superalloy as a function of the thermal parameter G- 1/2»R~ 1/4 42 ' 43. 33

3. RESEARCH OBJECTIVES A literature review revealed that some solidification parameters, such as dendrite arm spacing, have been investigated in titanium alloys and that a formation mechanism of beta fleck has been proposed. However, the proposed mechanism is not convincing since no systematic study has been conducted to verify the mechanism and there has been little data reported on the parameters, which is necessary to validate the mechanism. As liquid metal flow at the liquid/solid interface, which might lead to the formation of beta-fleck, is taken into consideration, permeability of the liquid metal through dendrite structure has to be discussed. Therefore, dendrite arm spacing is a critical parameter and may be used in criteria to determine if density-driven flow of liquid metal occurs during solidification by way of the basic Rayleigh number, as shown in (Eq.8). Particularly, it is critical to clarify the dendrite arm spacing in segregation sensitive titanium alloys, in which no effective etching method has been developed to make dendrite structure visible. A method to obtain dimensional and morphological information on dendrite structures, except for a conventional etching method, should be established in order to determine dendrite arm spacing in these titanium alloys. Densities of liquid and solid metal at the interface and the fraction solidified at the initiation of beta flecks should be clarified in considering the formation mechanism of betafleck. Liquid and solid metal density can be estimated from the chemical composition of beta-flecks with the calculation software, "Metals". The volume fraction of solid at the liquid/solid interface can be calculated by putting distribution coefficients into the Scheil equation, shown in (Eq.3). However, distribution coefficients of alloying elements in practical alloys containing multi-components have not been clarified yet. It is considered 34

to be difficult to assess the accurate volume fraction in these alloys by distribution coefficients obtained from the binary equilibrium phase diagrams. Therefore, the present study is focused on the following items as the research objectives : (1) To experimentally determine the dendrite arm spacing in segregation sensitive titanium alloys. The experimental methodology to obtain dimensions of dendrite arm spacing will be established by applying EPMA and the relationship between dendrite arm spacing and solidification conditions is planned to be clarified. (2) To determine distribution coefficients in practical titanium alloys consisting of multicomponent system. The segregation behavior of alloying elements in practical alloys has to be investigated under solidification conditions close to the equilibrium state. (3) To establish the formation mechanism of beta fleck by applying parameters obtained from (1) and (2). Finally, the validity of the proposed model must be examined by applying dendrite arm spacing and distribution coefficients, which are experimentally obtained, to the Rayleigh number. 35

4. EXPERIMENTAL METHODOLOGY 4.1 CHOICE OF ALLOYS The following three alloys have been selected for this experimental investigation. 1. 10-2-3 (Ti-10%V-2%Fe-3%Al) 2. Ti-17 (Ti-5%Al-2%Sn-2%Zr-4%Mo-4%Cr) 3. 6242 (Ti-6%Al-2%Sn-4%Zr-2%Mo) All three of these alloys are industrially used titanium alloys. 10-2-3 and Ti-17 are beta titanium alloys, both of which are used as component materials in airplanes, especially landing gear, etc 1-44. 10-2-3 and Ti-17 contain a large amount of beta stabilizing elements, such as 10 wt.% of vanadium and 2 wt.% of iron in the former and 2 wt.% of zirconium, 4 wt.% of molybdenum and 4 wt.% of chromium in the latter in order to stabilize beta phase at room temperature. In practice, however, beta phase is metastable at room temperature and alpha phase precipitates during cooling from the beta phase region with a slow cooling rate. It is expected that iron may segregate heavily in 10-2-3, while severe segregation of chromium would occur in Ti-17 during solidification. On the other hand, 6242 is an alpha+beta alloy, more precisely a near-alpha alloy, which is mainly used for high temperature services, such as compressor section's components of aircraft gas turbine engines 1-44. 6242 contains a higher concentration of aluminum and lower concentrations of beta stabilizing elements to stabilize both alpha and beta phase from room temperature to operating temperature. In this alloy, molybdenum is assumed to segregate heavily during solidification. Considering each alloying element in these alloys, aluminum is an alloying element 36

common to all three alloys; molybdenum and zirconium are contained in Ti-17 and 6242. Therefore, a difference in segregation behavior of the above alloying elements can be examined in different alloying systems as well as that depending on a difference between binary system and multi-component system. 4.2 EXPERIMENTAL METHODS Segregation behavior of iron in 10-2-3 small ingots melted and cast in an argon arc melting furnace In order to clarify the relationship between the solidification conditions and the segregation behavior of iron together with microstructures, 10-2-3 was melted using an argon arc melting furnace in the Advanced Materials and Process Engineering Laboratory (AMPEL) at The University of British Columbia(UBC). A picture of the furnace and its schematic layout are shown in Figures 14 and 15, respectively. During melting experiments, the chamber was kept in an argon gas atmosphere with a pressure of 35kPa in order to protect the molten metal from air. Typical operation conditions during melting were 10-15 volts and 1100-1500 amperes. The shape of molten metal pool was monitored through sight glasses during the experiments. Molten metal was superheated for 2-5 minutes, while the graphite crucible was preheated with an induction coil. After the molten metal was superheated, the water-cooled copper bottom plate was pulled out and molten metal was poured into the mold. Dimensions of crucible were 20 mm in inner diameter and 60 mm in length, if the crucible was filled with molten metal. Ingots were cut in half in the longitudinal direction and the transversal surfaces were polished to 1 diamond and finally etched for optical microscopic observation with an acid solution containing HF:HN03:H20(15 ml:50 ml:535 ml). Microstructures were observed with an 37

optical microscope. The segregation behavior of iron was analyzed by Electron Probe Micro Analysis (EPMA) attached to a Hitachi S-570 Scanning Electron Microscope (SEM) in the Department of Metals and Materials Engineering at UBC. The acceleration voltage of electron beam was 20kV and the analysis was carried out under a vacuum of lxlchpa. Segregation behavior of iron was examined in the directions inclined from the horizontal line by 0, 30, 45 and 60 in a cut plane to clarify the direction of dendrite arm. The intensity of Fe-Ka were counted and calibrated to weight per cent. Time evolution of temperature in the ingots during solidification was monitored in the argon arc furnace casts. Type-C thermocouples (tungsten-5wt.%rhenium vs. tungsten- 26wt.%rhenium) with wire thicknesses of 10/1000"(0.25 mm), inserted into alumina tubes for protection, were set in the vertical direction at 25 mm and 15 mm in height from the bottom of the crucible. Thermocouples were connected to a laptop computer by way of the temperature acquisition system "instrunet'. Temperature was read and recorded by the "instrunet' software installed on the computer. The frequency for reading temperatures (voltage) was 5Hz. Segregation behavior of alloying elements in production ingots In order to establish a method to determine dendrite arm spacing in production titanium alloy ingots, chemical analysis of alloying elements was carried out on specimens cut from a Ti-17 ingot and a 10-2-3 ingot with EPMA. The Ti-17 ingot was supplied by Timet Corporation and had dimensions of 450 mm in width and 430 mm in thickness. A 160 mm x 350 mm x 20 mm plate cut from a 10-2-3 ingot was supplied by RMI Company, which had initial dimensions of 760 mm in diameter and 2164 mm in length 27 >. Specimens 38

with a cross-section of 10 mm x 10 mm were cut from the sample materials and the transversal surfaces were polished to 6 ^m diamond. The segregation behavior of chromium in Ti-17 and that of iron in 10-2-3 were analyzed by EPMA attached to a Hitachi S-570 SEM as mentioned above. The acceleration voltage of electron beam was 20 kv and analysis was carried out under a vacuum of lxlo" 4 Pa. Analysis was conducted in the direction assumed to be perpendicular to the solidification direction, to investigate the distribution of chromium in Ti-17 and that of iron in 10-2-3. The intensities of Cr-Ka and Fe-Ka peaks were counted and calibrated to wt.%. Segregation behavior of alloying elements in laboratory melted small ingots using a zone melting furnace 10-2-3, Ti-17 and 6242 (Ti-6%Al-2%Sn-4%Zr-2%Mo) were melted and cast in a unidirectional induction levitation furnace at the Wright Patterson Air Force Laboratory in Ohio. Bar samples were heated and melted with a 30 mm long induction coil moving along the longitudinal direction. The coil moving speed was 4 mm/hr (l.llxlo- 6 m/sec) and the maximum temperature during melting was held at TL+20-30 K (TL: the liquidus temperature), which corresponds to a temperature gradient of 5.5-5.7xl0 4 K/m. During the experiments, samples were shrouded with argon gas which not only protected the system from oxidation, but also minimized elemental loss due to evaporation. Evaporation is a particular problem in the alloy systems chosen and for example negated the choice of levitation electron beam zone refining as a possible method for this experiment. Dimensions of the samples were 0.5"(12.7 mm) in diameter and 5"(127 mm) in length. The samples were cut in half in the longitudinal direction initially and then cut into three 39

specimens containing the start, the middle and the finish of melting. The specimens were polished to 1 um diamond and finally etched for optical microscopic observation. As an etchant, an acid solution containing HF:HN03:H20(15 ml:50 ml: 435 ml) was used for 6242 and Ti-17, while that consisting of HF:HN03:H 2 0(15 ml:50 ml: 535 ml) was used for 10-2- 3. After microstructural observation, specimens were re-polished to 6pm diamond before chemical analysis. The segregation behavior of the alloying elements was analyzed using an energy dispersion spectrometer (EDX) microprobe (KEVEX detector and a Quartz Xone analyzer) attached to a Hitachi S-570 SEM at UBC. Acceleration voltage of the electron beam was 20 kv and the analysis was carried out under a vacuum of lxlo -4 Pa in the longitudinal direction of each sample. The intensity of Al-Ka, V-Ka, Cr-Ka, Zr-Ka, Mo-La and Sn-La peaks was measured in each alloy and calibrated into wt.%. 40

Figure 15 Argon arc melting furnace in AMPEL, UBC. 41

Movable tungsten electrode I Sight glass Tubes for cooling water Argon gas c * c Molten metal Arc 0 3 Pressure Gauge Removable copper bottom plate Tubes for cooling Water IH coil Alumina container Graphite crucible Thermocouples Tubes for Cooling water Connected to a PC Figure 16 Schematic diagram of the argon arc melting furnace in AMPEL. 42

5. EXPERIMENTAL RESULTS 5.1 SEGREGATION BEHAVIOR OF ALLOYING ELEMENTS IN TITANIUM ALLOY INGOTS SOLIDIFIED IN DENDRITIC MANNER Segregation behavior of iron in laboratory melted 10-2-3 ingots using an argon arc furnace The melting experiments were conducted in an electrode arc furnace using an argon atmosphere. A total of six charges were made and four ingots were obtained. In one of the cases, an ingot was not obtained because of excessive oxidation during the experiment due to a shortage of argon gas supply and in another case, an ingot was not obtained due to failure of the electrode during melting. The results for each charge, including the two failed attempts, are summarized in Table 3. Table 3 Melting experiment results with the argon arc furnace Cast Ingot Holding time before Microstructure Notes no length pouring the molten metal 1 50mm 1 minute Equiaxed Initial conditions 2 45mm 2 minutes Partially elongated More superheat than No.l 3 Failed N/A N/A Excessive oxidation in molten metal 4 50mm 3 minutes Mostly elongated More superheat than No.2 5 Failed N/A N/A Electrode dropped during melting 6 50mm 5 minutes Equiaxed More superheat than No.4 Micrographs of the etched ingot samples 1,2,4 and 6 are shown in Figure 17(a)-(d), respectively. The microstructures listed in Table 3 were categorized from the results in the shape of the prior-beta grains, as shown in the figures. The equiaxed prior-beta grains were observed mainly in samples 1,2 and 6, while elongated grains along the longitudinal direction of the ingot could be seen in sample 4(Figure 17(c)). In general, dendritic 43

columnar growth is promoted under solidification conditions with a higher temperature gradient and a lower solidification velocity 5, etc. More superheat in the molten metal and more preheat into the crucible was applied by increasing the holding time before pouring the molten metal into the mold in sample 4 than in sample 1 or 2. This might have caused the elongated prior-beta grains. The holding time was prolonged to 5 minutes in sample 6, 2 minutes longer than in sample 4. The microstructure of sample 6, however, consisted of the equiaxed prior-beta grains, which were distributed uniformly. The iron concentration distribution in the directions inclined from the horizontal line by 0,30,45 and 60 0 on a cut surface in sample 4 ingot are shown in Figures 18-21. Both of the average of iron concentration and the statistical error of the data were deviated depending on the analyzed direction: from 1.589 to 1.780 wt.% in the former and from 0.083 to 0.113 in the latter. In the figures, error max shows a summation of the average concentration and statistical error, while error min shows a subtraction from the average concentration by statistical error. Among the figures, the concentration curve obtained from the horizontal direction showed the clearest peaks and periodicity in the curve, presenting almost the same spacing of 47 um between the peaks. In other figures, in the directions inclined from the horizontal line by 30-60, a difference between the peak and bottom concentration decreased but showed periodicity and a spacing of 42-48 um between the peaks. Evolution of temperature with time at the height of 25 mm and 15 mm from the bottom of a CP titanium ingot is shown in Figure 22. There were not enough raw materials of 10-2-3 and CP titanium was used in these experiments, whose thermal conductivity is close to that of 10-2-3. In this experiment, the holding time before pouring molten metal into the mold was 3 minutes and the induction heating system was switched off just prior to 44

pouring. This condition was very close to that applied for melting sample 4 ingot. In the figure, temperature rapidly increased at 7 seconds after slowly decreasing from 1323 K (1050 C) for the 25mm location and from 1123 K (850 C) for the 15 mm location, which indicates that the molten metal filled the crucible and touched the thermocouples at 7 seconds. Faster response in a temperature curve at 15 mm than that at 25 mm shows that molten metal piled up steadily from the bottom of the crucible, indicating that temperatures were monitored satisfactorily. Cooling rates were different depending on the location of thermocouples and the extent of solidification. From 1923 K (1700 C) to 1873 K (1650 C), the cooling rates were 13.2 K/s at 25 mm and 47.0 K/s at 15 mm, respectively. A clear change in cooling rate at around the melting point was not observed at either location. Segregation behavior of alloying elements in production ingots A photograph showing the transverse cross section of a Ti-17 ingot and the location of specimens taken for chemical analysis with EPMA is shown in Figure 23. Since no effective etching technique has been developed for titanium alloys, which allows us to identify the dendritic structure directly, the degree of the columnar growth of dendrites can only be estimated from the shape of the prior-beta grains. Microstructure of the cross section of the ingot consisted of elongated prior-beta grains originated mainly from the edges toward the center, which might be the traces of the growth direction of the dendrites. Specimen No.4, the closest to the center of the ingot as shown in Figure 23, was eventually used for the analysis and consisted mainly of the equiaxed prior-beta grains. An optical micrograph of Specimen No.4 is shown in Figure 24. The microstructure consisted of the equiaxed prior-beta grains andfinelydistributed platelet alpha and beta 45

grains in each grain. Chemical analysis with EPMA on chromium content was carried out in the two diagonal directions of the sample, which are assumed to correspond to the directions perpendicular and parallel to the elongated direction of the prior-beta grains, respectively. The distribution of chromium concentration obtained by EPMA in the direction perpendicular to the elongated direction of the prior-beta grains is presented in Figure 25. It is to be noted that the ingot used was an experimental one and did not have the conventional Ti-17 composition 45-46 of 4 wt.% and the average chromium concentration was 2.295wt.%. Error max and error min in the figure were obtained from calculation between the average concentration of chromium and a statistical error in chromium concentrations. As can be seen in Figure 25, the distribution of chromium concentration has periodicity consisting of 3 peaks, 2.66, 2.59 and 2.62 wt.% of chromium contents and shows the spacing of 1062 and 1593 um between the peaks. These three concentration values at peaks correspond to 1.13-1.16 times as high as that of the average and much higher than the maximum error value of 2.42 wt.%. It is therefore clear that concentration of chromium fluctuated periodically in the direction perpendicular to that of solidification. Diffusion of solute chromium atoms in the solid state after solidification was taken into consideration according to (Eq.4) and (Eq.5) 19-25 in the production Ti-17 ingot. Calibrated chromium concentrations are shown in Figure 26. A solid line indicates the chromium concentrations measured by EPMA, while a broken line shows the chromium concentrations just after the solidification, which was estimated from considering the effect of diffusion of chromium atoms in the solid state. Diffusion of solute chromium atoms in the solid state had little effect on a change in concentration, which caused an increase of only 0.02 wt.% (from 2.67 to 2.69 wt.%). 46

The segregation behavior of iron in a 10-2-3 production ingot was investigated. Figure 27 shows a macrograph of an as-received 10-2-3 plate material. A 30" diameter ingot was cut into 1" thick plate material in the longitudinal direction of the ingot 27 and one surface of the material was polished and etched. The microstructure consisted of the equiaxed prior-beta grains, with grain diameters varying from 5 to 15 mm, distributed uniformly throughout the material. Block samples (sample No. RIO 1-104) were cut from the marked locations in Figure 27. The distribution of iron concentration obtained in the direction inclined to the horizontal direction by 60 in the R102 sample is shown in Figure 28. In the analyzed area, three sharp peaks can be observed with spacing of 1062 and 1151 um between the peaks, indicating periodicity as seen in a Ti-17 production ingot. Clear peaks and periodicity of concentration profiles could not be seen in any other directions. In order to clarify if beta-flecks exist in this area, EPMA analysis for iron concentration was conducted on each of these four samples. The distribution of iron concentration is shown in Figure 29, in which the numbers indicate iron concentration in weight percent in each section. The maximum iron concentration measured was 1.89 wt.%. This value is not so different from the average concentration of 1.67 wt.%, showing that beta-flecks did not exist in this region. 47

5.2 SEGREGATION BEHAVIOR OF ALLOYING ELEMENTS IN ZONE MELTED TITANIUM ALLOY INGOTS Photographs of as-received specimens cast with a zone-melting furnace are shown in Figure 30(a)-(c). The 6242 specimen consisted of one full-length bar but the Ti-17 and 10-2-3 specimens were separated into two parts. The full length of the specimens was about 200 mm and that of the melted part was 70-80 mm. A step-like expanded shape could be seen at the start point of melting, where the diameter was 1-2 mm larger than that of the unmelted part. It may have arisen because melting was started at a more downward location than the middle of the specimen, corresponding to 70 mm from the bottom end, and it proceeded upward in the vertical direction. Thefinalpoint of melting resulted in a necked shape, where the Ti-17 and 10-2-3 specimens were separated into two parts. However, the shape of the fractured surface was quite different for the Ti-17 and 10-2-3 specimens. A fibrous and zigzag surface was seen in the Ti-17 specimen, showing the typical ductile fractured surface, which might have occurred at around room temperature. It is considered that the specimen was separated into two parts after solidification had been completed. In contrast, a smooth surface and spherical shape was seen on the separated end in the 10-2-3 specimen, which shows that the fractured section remelted. In analyzing the chemical composition of the 10-2-3 specimen, this effect has to be taken into consideration. The cross-sectional macrographs of the Ti-17, 10-2-3 and 6242 specimens are shown in Figures 31-33. In all the samples, the unmelted parts consisted of fine equiaxed grains close to both ends, which is the initial microstructure of the materials. At locations closer to the start orfinishpoint of melting from each end, the grain diameter becomes coarser, 48

showing grain growth occurred by the heat input during the melting experiment. At the middle section, no clear grain boundary was identified and the structure appeared a single crystal until the melting section terminated. There were subtle differences in the microstructures in each specimen. In the Ti-17 and 10-2-3 specimens, some localized areas around the tip of the final melting location was difficult to etch, indicating the chemical compositions of these areas may be considerably different from those of the other parts (Figures 31 and 32). In the middle part of melting in the 10-2-3 specimen, unclear lines streaked in the longitudinal directions, which might be subgrains (Figure 32). Two different etched colors or patterns can be seen in the middle part of the 6242 specimen but no obvious grain boundaries were found between the different patterns. Detailed observation revealed that these different etched patterns depended on the degree of etching and they consisted of an acicular microstructure with the same configuration and direction of laths. t The distribution of alloying elements analyzed with EDX in the longitudinal direction of the 10-2-3, Ti-17 and 6242 specimens are shown in Figures 34-36. These figures were obtained by combining data from three specimens cut in the longitudinal direction for each alloy. The concentration distribution of each alloying element showed a reasonably smooth shape. However, the actual concentration profile obtained for the 10-2-3 specimen was not smooth, as shown in Figure 37. For example, around the final melting point, two irregular peaks were seen for the iron concentration profile in the figure, comparing with a smooth curve with a peak for chromium in Ti-17 or for zirconium in 6242. It is thought that this resulted from the redistribution of alloying elements due to remelting after the sample had been separated. Therefore, paying attention to the similar concentration profiles of alloying elements at the x-axis at 71-73 mm and at 82-84 mm, concentration 49

data at 73-82 mm were rejected. The final results are presented in Figure 34. The concentration profiles should have been the one shown in Figure 34, if the samples had not remelted. Figures 34-36 show similar segregation behavior of other segregating elements: iron in 10-2-3, chromium and zirconium in Ti-17 and zirconium in 6242, all of which decreased as melting started and increased rapidly near the finishing point. The maximum content of iron close to the end of melting in 10-2-3 reached to 4.56 wt.%, which corresponded to 2.73 times the average iron content of the parent metal, 1.67 wt%. Brooks et al 7 reported that beta-flecks formed in regions containing higher than 2.4 wt.% of iron and it is surmised that beta fleck formed close to the final melting point in this sample. In Ti-17, a large increase in chromium content was identified at the final melting point, the increase of which was 1.83 times higher than the average. Concentrations of elements other than iron, chromium and zirconium indicated reverse profiles, showing an increase at the start and a decrease near the finish. The molybdenum content showed curious behavior, a much higher decrease in Ti-17 than in 10-2-3 at the final melting point. It shows that even the same alloying element can have different concentration profiles in different alloying systems. The broken lines in Figures 34-36 represent the concentration vs. is (fraction solidified) curves obtained using the Scheil equation (Eq.(3)) for iron in 10-2-3, chromium in Ti-17 and zirconium in 6242, respectively. f s was calculated as a ratio between the distance from the initial melting point and the distance of the melted section. The actual concentration profiles obtained from experiments have a steeper shape than the curves determined by the Scheil equation. In reference, concentrations of oxygen and nitrogen in a 10-2-3 ingot produced in the same zone-melting furnace are presented in Figures 38 and 39. These measurements were 50

obtained from LECO analysis conducted by TIMET Corporation 47. The concentration profiles of oxygen and nitrogen show an increase at the start point and a decrease at the finish point, such as molybdenum, aluminum, etc. The concentration data of each alloying element obtained from the zone-melted 10-2-3, Ti-17 and 6242 alloy ingots is summarized in Tables 4-7. Each table contains the average (Cave), maximum(cmax) and minimum(cmin) concentrations, the ratio between the maximum and minimum concentration(cmax/cmin) and the segregation coefficient(k) for each alloying element. The average concentration corresponds to the average value of concentrations obtained in the unmelted parts, which is supposed to be the initial composition (Co) of each alloying element. The segregation coefficient was calculated on the basis that the concentration of each element at the start of melting (CL) equals the product of the segregation coefficient and the average concentration (k»co) 48. As can be seen in Figures 34-36, the concentration at melting start point corresponded to the maximum concentration for aluminum, molybdenum and tin, and the minimum concentration for iron, vanadium, chromium and zirconium. Therefore, the segregation coefficients of the former three elements were obtained from the ratio of Cmax/Cave, while those of the latter four elements were given by the ratio of Cmin/Cave. Equilibrium distribution coefficients (keq) of each alloying element, calculated from the binary phase diagrams 20, are included in the tables. In Table 4, Cmax/Cmin of iron is very high, 7.23, which shows that iron segregates heavily in 10-2-3. Cmax/Cmin ratios of tin, zirconium, molybdenum and chromium in Ti-17 and that of zirconium in 6242 are higher than 2, indicating these elements also segregate heavily in each alloy. On the other hand, aluminum in all the alloys, vanadium, oxygen and nitrogen in 10-2-3 and tin and molybdenum in Ti-17 show Cmax/Cmin ratios smaller 51

than 2, indicating that they do not segregate heavily in these alloys. Tin shows interesting behavior; it shows segregation increased in Ti-17 with a Cmax/Cmin ratio of 2.06 versus 1.30 in 6242. This result indicates that the same alloying element can show differences in the degree of segregation in different alloying systems. 4 < The segregation coefficients of some alloying elements are different depending on alloying systems. For instance, k of aluminum fluctuates from 1.02 in 6242 to 1.13 in 10-2- 3, while that of tin varies from 1.08 to 1.15, etc, although the difference between them is not so large. A large difference between k and k eq can be seen in iron in 10-2-3, tin, zirconium, molybdenum and chromium in Ti-17 and tin and zirconium in 6242. In particular, k is much less than k e q for iron in 10-2-3. 52

Table 4 Composition variations and segregation coefficients(k) of alloying elements in a zone melted 10-2-3 ingot Alloying Cave Cmax Cmin Cmax k keq(ref(20)) Element wt.% wt.% wt.% /Cmin Al 2.76 3.11 2.35 1.32 1.13 1.05 V 9.92 10.53 9.15 1.15 0.95 0.95 Fe 1.67 4.56 0.63 7.23 0.38 0.60 Table 5 Composition variations and segregation coefficients(k) of alloying elements in a zone melted Ti-17 ingot Alloying Cave Cmax Cmin Cmax k k eq (ref(20)) Element Wt.% wt.% wt.% /Cmin Al 5.55 5.86 4.80 1.22 1.06 1.05 Sn 1.52 1.75 0.85 2.06 1.15 0.92 Zr 1.39 2.53 1.07 2.36 0.77 0.90 Mo 2.75 3.16 1.30 2.43 1.15 1.50 Cr 5.50 10.04 4.06 2.47 0.74 0.70 Table 6 Composition variations and segregation coefficients(k) of alloying elements in a zone melted 6242 ingot Alloying Cave Cmax Cmin Cmax k k eq (ref(20)) Element Wt.% Wt.% Wt.% /Cmin Al 5.83 5.94 5.66 1.05 1.02 1.05 Sn 1.50 1.62 1.25 1.30 1.08 0.92 Zr 2.63 4.49 1.90 2.36 0.72 0.90 Mo 1.34 1.46 1.03 1.42 1.09 1.50 Table 7 Composition variations and segregation coefficients(k) of oxygen and nitrogen in a zone melted 10-2-3 ingot(ref(47)) Alloying Cave Cmax Cmin Cmax k k eq (ref(20)) Element wt.% wt.% wt.% /Cmin 0 0.137 0.182 0.118 1.54 1.33 1.60 N 0.011 0.016 0.009 1.78 1.45 1.58 53

5.3 DENSITY OF BETA-FLECK AND LIQUID METAL CALCULATED USING "METALS" FOR 10-2-3 AND Ti-17 ALLOYS The density of beta-fleck and liquid metal was obtained by calculation using "METALS", the principles of which are shown in Appendix A. Examples of the calculated results are presented in Figures 40-43. In Figure 40, the density of solid phase was calculated in an iron concentration range from 0.5 to 2 wt.% at 1900 K, while that of the liquid phase was calculated in a range from 2 to 5 wt.% at 1905 K(the melting temperature). These ranges were determined on the basis of the zone-melting experiment results, where the minimum iron concentration was 0.63 wt.% and the maximum iron concentration was 4.56 wt.%. There is a wide gap between the density of solid and liquid phases, with the liquid showing lower density than the solid for the same composition at a similar temperature. For instance, the density of the solid phase containing 2 % iron is 4497.8 kg/m 3 at 1900 K, while that of the liquid phase containing 2 % iron is 4164.8 kg/m 3 at 1905 K. The density increases with iron concentration, however; even though the liquid phase contained 5 % iron, it is still lighter than solid phase consisting of 0.5 % iron. The density of liquid, which might form beta-flecks in 10-2-3, is estimated to be higher than 4171.6 kg/m 3 at the melting temperature since the iron concentration at beta-fleck is at least 3.1 %. The effect of temperature on density of the liquid(beta-flecks) and solid phase is shown in Figure 41, where the calculation was conducted for liquid containing 3.1 % iron and the solid containing 2.0 % iron. It is clear from Figure 41 that the density increases rapidly as the alloy solidifies. The density increases with a decrease in temperature in the solid and liquid phase but the slope is steeper in liquid than in solid. 54

Calculated results obtained for Ti-17 are shown in Figures 42 and 43. The density of liquid metal containing 5.5 % chromium, which is known as a composition of beta-flecks in Ti-17 13, was higher than 4167.9 kg/m 3 at melting temperature (1914 K), while that of solid phase containing 4% chromium was 4503.4 kg/m 3. In Ti-17, the density of the solid is heavier than that of the liquid, which is the same trend as observed in 10-2-3. 55

(a) Sample 1 (b) Sample 2 Figure 17 Macrographs of 10-2-3 ingots melted by the argon arc furnace (continued). 56

(c) Sample 4 (d) Sample 6 Figure 17 Macrographs of 10-2-3 ingots melted by the argon arc furnace. 57

3 2.5 Ti-10%V-2%Fe-3%AI cast ingot Ingot diameter: 20mm, length : 50mm Analyrert hy FPMA (Fe-Kry) Accelerated Voltage : 20kV Fe(ave)=1.780wt.%, cr=0.113 Ingot bottom r/2 ddeg t/3! -*-pitch=14.2# m - - error max -a- error min Ingot Top 200 300 400 500 distance from the start point (nm) 600 700 Figure 18 Distribution of iron concentration in the horizontal direction in a 10-2-3 laboratory-melted ingot. 200 400 600 800 1000 1200 distance from the start point (u m) ure 19 Distribution of iron concentration distribution in the direction inclined to the horizontal direction by 30 in a 10-2-3 laboratory-melted ingot. 58

Ti-10%V-2%Fe-3%AI cast ingot Ingot diameter: 20mm, length : 50mm Analyzed by EPMA (Fe-Ka) Ingot bottom Itigot top -pitch=14.2j/ m - error max - error min 0.5 200 400 600 800 distance from the start point (u m) 1000 1200 Figure 20 Distribution of iron concentration in the direction inclined to the horizontal direction by 45 in a laboratory-melted 10-2-3 ingot. 2.5 Ti-10%V-2%Fe-3%AI cast ingot Ingot diameter: 20mm, length : 50mm Analyzed by EPMA (Fe-KoQ Accelerated Voltage: 20kV Fe(ave)=1.594wt.%, a =0.083 Ingot bottom -pitch=14.2# m - error max - error min Ingot top 0.5 100 200 300 400 500 600 700 distance from the start point (u m) 800 900 1000 Figure 21 Distribution of iron concentration in the direction inclined to the horizontal direction by 60 in a 10-2-3 laboratory-melted ingot. 59

20 25 time (sec) Figure 22 Evolution of temperature with time during solidification in Commercially Pure Titanium melted in an argon arc furnace. 60

500Lim Figure 24 Microstructure of a Ti-17 production ingot. 61

3.2 2.8 Ti-17 (Ti-5%AI-2%Sn-2%Zr-4%Mo-4%Cr) as cast slab Slab width: 450mm, thickness: 430mm Analyzed by EPMA (Cr-Ka) Accelerated Voltage = 20kV Cr(ave)=2.295wt.%, CT=0.120 1062U m 1593 um 1.8 -chromium concentration - error max -error min 1.6 500 1000 1500 2000 2500 distance from the sample edge (u m) 3000 3500 4000 Figure 25 Distribution of chromium concentration in a Ti-17 production ingot. 3.2 2.8 Ti-17 (Ti-5%AI-2%Sn-2%Zr-4%Mo-4%Cr) production ingot Slab width: 450mm, thickness: 430mm Analyzed by EPMA (Cr-Ka) Accelerated Voltage = 20kV Cr(ave)=2.295wt.%, <r=0.120 2.6 8 2.4 2.2 1.8 - measured chromium concentration just after solidification 1.6 500 1000 1500 2000 2500 distance from the sample edge (u m) 3000 3500 Figure 26 Estimated distribution of chromium concentration in a Ti-17 production in just after solidification. 62

1 2 3 4 5 6 7 8 9 im : j rih11, t h j If I r 1 1 j1111.;11. j l i!11 s f11 i tl I; IL :^.11, f11 i r11 f? f.h: j l;111, r! I'11 r I! t r;: i; n11 i 1 i111 11 12 13 14 15 16 17 18 19 21 Figure 27 Macrograph of as-received Ti-10-2-3 production ingot. 2.5 2.25 Ti-10-2-3 (Ti-10%V-2%Fe-3%AI) production ingot Ingot diameter:760mm, length:2413 mm Analyzed by EPMA (Fe-Kar) -Accelerated Voltage = 20kv Cr(ave)=1.665wt.%, a =0.087 1.25 1062jim 1150.5^m + * < - measured data - error max - error min 1000 2000 3000 4000 5000 distance from the sample edge (u m) 6000 7000 Figure 28 Distribution of iron concentration in the direction inclined to horizontal direction by 60 in a 10-2-3 production ingot. G3

Sample No RM101 RM102 RM103 RM104 iron concentration (wt%) 1.67 1.59 1.61 1.47 1.63 1.71 1.80 1.85 1.55 1.86 1.51 1.73 1.59 1.70 1.66 1.43 1.65 1.75 1.64 1.78 1.56 1.83 1.76 1.54 1.74 1.65 1.74 1.61 1.76 1.88 1.50 1.66 1.57 1.77 1.59 1.45 1.88 1.71 1.61 1.55 1.63 1.65 1.52 1.89 1.58 1.57 1.65 1.56 1.69 1.66 1.77 1.68 1.69 1.84 1.53 1.62 1.84 1.82 1.86 1.50 1.59 1.68 1.50 1.58 Figure 29 Distribution of iron concentration in a 10-2-3 production ingot. 64

1 8 Direction of solidification < _ 9 11 12 13 14 15 16 17 1 9 I!) 21 22 23 24 25 26 (a) 10-2-3 5 6 7 8 (TJ 11 12 13 H 15 '8 " 19 21 22 23 24 25 26 (b) Ti-17, j.m-r- 5 6 7,8. 9 0 21 22 23 24 25 8» 11 12 13 1* 15 16 17 ii III imiaii^imiintirtntift I I.Ij iilwlbm#!.t (c) 6242 Figure 30 Macrographs of zone melted samples (as-received). 65